Re: Deduplicating __NR_* and SYS_* syscall defines

[Rich Felker <dalias@libc.org>]

On Fri, May 13, 2016 at 12:15:48PM +0200, Szabolcs Nagy wrote:
> * Rich Felker <dalias@libc.org> [2016-05-11 20:57:12 -0400]:
> > On Wed, May 11, 2016 at 07:22:30PM -0500, Bobby Bingham wrote:
> > > On Wed, May 11, 2016 at 05:53:12PM -0400, Rich Felker wrote:
> > > > On Mon, May 09, 2016 at 07:26:37PM -0500, Bobby Bingham wrote:
> > > > > During the powerpc64 review, Rich mentioned wanting to replace the
> > > > > arch/*/bits/syscall.h files with .in files that would be preprocessed with
> > > > > something like:
> > > > > 
> > > > >     sed -e p -e s/__NR_/SYS_/ < $< > $@
> > > > > 
> > > > > This would eliminate a lot of intra-file duplication here.
> 
> i don't know how much is it worth to keep the syntax consistent
> with kernel headers, i'd just clean these up with special syntax
> like TYPEDEF in alltypes e.g.
> 
> SYS write 1

This does save some space (maybe 25-30% size reduction?) but I think
it also comes at the sense of some flexibility and readability. For
example if there are ever archs where the definitions need to depend
on some compile-time macros (e.g. if we had an arch where it makes
sense to put a couple ABI variants with different syscalls together as
subarchs rather than independent archs) there becomes a question of
how to format those conditions and whether their directives would be
misinterpreted. Having #define on every line so that the file as-is is
valid preprocessor input avoids any such issues.

Of course these things are also a reason to possibly prefer the form
with SYS_ and __NR_ interleaved (my original sed command). With them
separate like the current proposed patch, anything inside a
conditional block or dependent on interleaved redefinition of other
macros would break.

> with consistent whitespace, removed comments etc.

Agree on consistent whitespace (and we should do that regardless of
the rest of this proposal), but note that it can't necessarily be just
3 space-delimited tokens. Some archs have arithmetic expressions for
the integer field, and requiring that not to contain whitespace would
be rather arbitrary and confusing, so the last field would probably
need to be the whole remainder of the line.

Rich